Artificial intelligence seemed more like science fiction than reality to many people just a few years ago. Now, it is everywhere, and soon, it could be making decisions on whether you qualify for a mortgage.
“We’re really most of the way there already,” says Noah Kenney, founder and principal consultant at Digital 520, a strategic technology advisory. “Rocket Mortgage and Better will approve you in eight minutes now.”
Lenders at the leading edge of AI are deploying chatbots to walk homebuyers through applications online and use specialized models to make decisions. Others are taking a more cautious approach and using AI to streamline tasks on the back end while keeping interaction with a human loan officer front and center.
[Read: Best Mortgage Lenders]
How Lenders Are Using AI
If and how lenders are using artificial intelligence may depend on what you consider AI, something Kenney says is not consistently defined in the United States.
Fannie Mae and Freddie Mac have used automated underwriting systems for more than 30 years to review applications for government-backed mortgages. However, those early systems were algorithms that determined if applications met certain criteria. They didn’t have the ability to analyze as today’s AI does, although those features are coming.
In January 2026, Fannie Mae updated its Desktop Underwriter tool to add machine learning models — which use AI — that are expected to better assess an applicant’s risk profile.
Online lenders have been at the forefront of using AI in the underwriting process. Rocket Mortgage announced in 2024 that it was launching Rocket Logic, a patented AI-driven technology that uses proprietary data and call transcripts to help close loans faster.
Last year, the company posted on its blog that it was working on a proprietary origination system and “An agent-powered conversational platform using multiple LLMs, designed to guide clients seamlessly through their journey.”
Traditional lenders, such as banks, seem to be taking a more conservative approach to using AI.
“We are using it to basically help put together the picture of complex tax returns,” says Brian Grzebin, president of mortgage banking at Univest Bank and Trust, which has branches in the mid-Atlantic. AI might extract and summarize data from lengthy and complex tax returns, but “Obviously, a human is making the decision.”
Most Homebuyers Still Want Human Touch
Most consumers see AI as an inevitable part of the homebuying process, according to a 2026 survey by Cotality, a financial services firm. Three in four buyers assume AI is already part of the process, with younger generations having more confidence in its benefits.
Although consumers expect AI will be used when buying a home, a majority — 55% — prefer to work with a human when it comes to securing a mortgage. That is up from 46% who said the same in a similar survey last year.
“Banks need to move thoughtfully,” says George Taylor, head of residential mortgage at M&T Bank, which services mortgages nationwide and has branches from Maine to Virginia. “There’s an opportunity for AI to streamline the process, but there needs to remain human involvement.”
Grzebin questions whether AI will ever be able to ask “poignant questions” that get to the heart of what people want and need. “It’s an emotional process,” he says of buying a home. “I don’t know that the AI will get to the point where (it) understands a person’s wants and needs. I don’t know if we will ever get a human out of the process.”
Pros of AI-Assisted Mortgages
There are three main benefits of using AI in the mortgage process, according to Kenney:
— Speed
— Cost
— Access
Getting approval from a human underwriter can take as long as 45 days using a traditional mortgage origination process. With AI, that time could be reduced to seconds.
Mortgage lender Better says its AI system can deliver decisions in as few as 45 seconds, with the median decision time at two minutes and 24 seconds. The company’s Tinman AI App uses OpenAI-owned ChatGPT to create a conversational credit decision engine that can be used by financial institutions and loan officers.
Meanwhile, it costs lenders about $11,600 to originate a loan, according to 2023 data from Freddie Mac, and labor accounts for about two-thirds of the cost. Using AI could reduce origination costs for lenders and, theoretically, for consumers too.
“AI models are incorporating more data,” Kenney says. That could help those with thin credit files and others who have traditionally encountered barriers when trying to access loans. “There’s a lot of evidence that they approve a lot of applicants in protected classes.”
Zest AI, which counts home equity loans among the products it serves, says its model increases loan approvals for several groups while keeping risk constant. In a 2024 press release, the company stated it had helped lenders increase approvals by the following percentages in these applicant groups:
— Latino: 49%
— Black: 41%
— Women: 40%
— Elderly: 36%
— Asian American and Pacific Islander: 31%
[See: Best Mortgage Lenders for First-Time Homebuyers]
Potential Drawbacks of Using AI for Mortgages
Although there is evidence that AI can improve approval odds for some loans, at least one study questions whether that is always true for mortgage lending.
Researchers at Lehigh University in Bethlehem, Pennsylvania, found that large language models — the foundational technology behind generative AI — consistently recommended denying loans and charging higher interest rates for Black applicants than for otherwise identical white applicants. The 2024 study found that the bias seemed to disappear when the LLM was instructed to disregard race.
“The models learn from historical data,” Kenney says. If that historical data is biased, it could lead to biased results in the future.
The Lehigh University study tested several AI models, such as ChatGPT, Anthropic’s Claude and Meta’s Llama. However, mortgage companies such as Rocket and Better are using proprietary AI models that may not have the same results.
“The AI is only as good as the information you put into it,” Grzebin says.
That’s one reason why he’s not sure mortgage lending will ever become a completely self-serve process. Borrowers may need the help of a human to understand what needs to be provided as part of an application, and some people may simply prefer to hand over documents and let someone else do the work of uploading or inputting information.
“We see AI as a tool that can streamline the process but not replace the human experience,” Taylor says.
Mortgage Regulations Apply to AI Models
As for transparency, the same rules apply regardless of whether a human or AI makes a lending decision.
“A lot of the regulatory framework already exists,” Kenney says. He points to federal quality control standards and data privacy laws as examples.
In 2023, the Consumer Financial Protection Bureau issued guidance stating that lenders must provide a specific reason for a denial even when using AI or other complex models. States may be looking at adding their own regulations. For instance, the Colorado AI Act was amended this year to require that consumers be notified in advance that “automated decision-making technology” could influence a decision and explain how data inputs will be used.
More from U.S. News
Kevin Warsh Set to Lead the Fed: Why a Trump-Backed Chair Doesn’t Guarantee Lower Mortgage Rates
What a Recession Would Mean for Mortgage Rates in 2026
Is a 15- or 30-Year Mortgage Right for You?
No Broker, No Bank, No Problem: How AI Could Turn Mortgages Into a Self-Service Experience originally appeared on usnews.com
